The notion of immunity to error through misidentification (IEM) has played a central role in discussions of first-person thought. Its claim to centrality rests, I think, on two main planks.
First, asking whether certain first-person thoughts are IEM has appeared to be a way of getting at the same kind of maximally deep philosophical issue as is raised by asking whether certain first-person judgments are about oneself as subject.Footnote 1 So, the interest of this latter question is inherited by claims about which first-person judgments are IEM. But the notion of IEM is reasonably precisely defined, in a way that that the notion of thought about oneself as subject is not. So, the obscurity of this question is not inherited by claims about which first-person judgments are IEM. IEM is an essential bit of philosophical apparatus because it allows us to make tractable a question that is of maximal philosophical interest but that, formulated as it previously had been, was impossible to settle.
Second, the distribution of IEM – which judgments have the property – has seemed to pattern in a way that connects strikingly with the distinction between mental and bodily properties of persons. The mental-bodily distinction has obviously played a huge role in the debate about the metaphysics of persons. But there is a separate question about the place the distinction has in the epistemology of first-person thought. Most philosophers who have discussed the distinction have thought that the only first-person judgments that are IEM in the fullest sense are mental first-person judgments. ‘I am in pain’, based on introspection, or ‘I am thinking that p’, based on introspection, will be IEM. ‘My knees are crossed’, based on proprioception, or ‘I am moving through a wooded grove’, based on vision, will not be. If they are right about that, then there is an epistemological asymmetry between our relationship to ourselves as bearers of bodily properties and our relationship to ourselves as bearers of mental properties (one which one might then use to try to establish metaphysical conclusions, or to diagnose metaphysical illusions, according to taste). If they are wrong about that, because both bodily and mental judgments can be IEM, then our bodily nature is deeply ingrained in the way we think about ourselves. One can think about oneself simultaneously as a subject, and as a bodily thing.Footnote 2
My discussion of IEM will focus on the following thesis:
Bodily first-person judgments, just like mental first-person judgments, can be IEM. (Parity Thesis)
The majority view, as just outlined, is that this thesis is false, and that it is false because bodily first-person judgments can’t be IEM in the fullest sense, whereas mental first-person judgments can be. I’ll argue that the two most influential notion attempts to craft a notion of IEM that supports this verdict fail and that, as things stand, we have been given no reason to reject the Parity Thesis.
Before getting into the main discussion, I take care of three preliminaries. First, I offer a quick recap of the debate about IEM thus far, highlighting the points relevant to the Parity Thesis. Second, I note that there are different notions of IEM, and I say why the Parity Thesis is best understood as a claim about the most demanding notion of IEM. Third, I say why no notion of IEM should be so demanding that it collapses into the notion of infallibility.
(i) Recap of the debate
The rejection of the Parity Thesis is clearly embedded in Wittgenstein’s original presentation of the distinction between uses of ‘I’ as subject and uses of ‘I’ as object. All of his examples of uses of ‘I’ as subject involve mental predications – ‘I see so and so’, ‘I hear so and so’, ‘I try to lift my arm’, ‘I think it will rain’, ‘I have a toothache’. All his examples of uses of ‘I’ as object involve bodily predications – ‘My arm is broken’, ‘I have grown six inches’, ‘I have a bump on my forehead’, ‘The wind blows my hair about’ (Wittgenstein Reference Wittgenstein1958, 66–7). Wittgenstein did not say why his examples exhibited the striking mental/physical pattern he took them too.
Shoemaker (Reference Shoemaker1968) introduced the idea of a quasi-memory to the discussion. He claimed that memory judgments are vulnerable to error through misidentification because they are vulnerable to quasi-memories, and quasi-memories, Shoemaker thought, involve a kind of misidentification. It just isn’t a kind of misidentification that de facto we have to worry about. Shoemaker’s notion of a quasi-memory generalizes to the notion of a ‘quasi-case’, which will be important in what follows. The generalization of Shoemaker’s suggestion about the significance of quasi-memories is that any judgment that is vulnerable to a quasi-case – and that, importantly, will include every bodily first-person judgment – will ipso facto be vulnerable to misidentification.
Gareth Evans (Reference Evans and McDowell1982, 241–8) defended the Parity Thesis against Shoemaker’s attack, arguing that, on a fairly flatfooted way of thinking about the notion of a misidentification that Shoemaker himself seemed to be operating with, quasi-cases do not involve misidentification.
The success of Evans’s defense has dictated the strategy of subsequent opponents of the Parity Thesis. Their strategy is to attempt to craft rather less flatfooted definitions of misidentification than Shoemaker’s on which quasi-cases do count as involving misidentification. Two main versions can be distinguished. The first invokes the notion of an undermining defeater. It is defended by James Pryor (Reference Pryor1999). The second centers on the notion of a background presupposition. It is developed first and in most detail by Annalisa Coliva (Reference Coliva2006, Reference Coliva, Evans, Prosser and Recanati2012), and has also been explored by Crispin Wright (Reference Wright, Prosser and Recanati2012). This paper argues that neither strategy is successful.Footnote 3
(ii) Different notions of IEM
Many theorists think, and I agree, that there are different kinds of IEM. Most theorists of IEM would say that it only the most demanding kind of IEM which bodily first-person judgments lack. For example, James Pryor writes:
‘There are several different sorts of error through misidentification, and they are importantly different. This paper aims to clarify and distinguish these different kinds of identification and misidentification and the correspondingly different notions of immunity to error through misidentification. I will argue that immunity to error comes in two main sorts … I will show how we can use these distinctions in assessing a debate between Shoemaker and Evans about whether first-person memory based judgments are immune to error through misidentification. Evans says they are immune. Shoemaker says they are not. I will argue that, with respect to the most interesting sort of immunity, Shoemaker is right. Our first-person memory judgments do not have that sort of immunity.’ (Reference Pryor1999, 272).
If there are several different notions of IEM, then there are different ways of reading the claim that bodily first-person judgments, just like mental first-person judgments, can be IEM. I intend the Parity Thesis to be understood as concerning the most demanding grade of IEM, whatever that turns out to be. It would be highly unimpressive to show that there is a grade of IEM that bodily as well as mental first-person judgment can have, while allowing that there is another that only mental first-person judgments can have. Against the background of the presumption that there is some interesting epistemic asymmetry between our relationship to our mental and our physical properties, it would be easy to insist in this situation that it is the more demanding grade of IEM that should be used to cash out the notion of thought about oneself as subject.
(iii) IEM and infallibility
On the other hand, I think that no adequate notion of IEM will be so demanding as to collapse into the notion of infallibility. This assumption does work in my argument. For example, I argue that Pryor’s attempt to define a notion of IEM cannot be right because, if it were, his notion would collapse into infallibility. What justifies this assumption that IEM and infallibility are different?
The first justification is that the collapse of one into the other would disqualify paradigm examples of IEM from being examples of the phenomenon. For example, ‘I am in pain’, based on introspection, is a paradigm example of an IEM judgment. However, many will accept that it is an exaggeration to say that ‘I am in pain’, based on introspection, is infallible. It certainly seems as though that one might introspectively mistake an itch for a pain (See Williamson Reference Williamson2000).
Not everyone will be convinced by this. Perhaps the mental is a domain in which the appearance/reality distinction doesn’t really arise. If a property is properly mental, someone might say, then one will be introspectively inclined to self-ascribe it only if one really instantiates it. In that case, ‘I am in pain’, made on the basis on introspection, will be infallible (and we will need some other way of thinking about what the pain-itch case shows). The notion of IEM’s collapsing into that of infallibility won’t cause paradigm examples of the phenomenon to be ruled out, so long as all paradigm examples of the phenomenon are infallible.
There is however a different justification for not allowing the collapse. If being IEM and being infallible turn out to be the very same epistemic status, latched onto in different ways, what is the point of introducing the notion of IEM in the first place? Any question that could be formulated using the notion of IEM could be formulated using the notion of infallibility instead. Since the notion of infallibility is older, and far simpler, introducing IEM would be a pointless complication. To that extent, accepting that IEM is a different epistemic status from infallibility is the price of admission to the debate about IEM. If they are the same status, then, on the face of it, the debate about IEM (qua IEM) shouldn’t exist.Footnote 4
Here is an outline of the argument that follows. Section (§1) sets out the notion of immunity to error through false-belief misidentification, and points out that it is an epistemic status that bodily first-person judgments can have. Section (§2) introduces the notion of a ‘quasi-case’, a kind of error-possibility to which every bodily first-person judgment seems vulnerable in principle. The remaining two sections focus on the two main attempts to craft a notion of IEM on which vulnerability to quasi-cases entails vulnerability to error through misidentification and on which Parity Thesis is therefore false. The first of these, discussed in section (§3), is Pryor’s. Pryor’s attempt focuses on immunity to error through wh-misidentification. I agree with Pryor that immunity to error through wh-misidentification is different from immunity to error through false belief misidentification. But I argue that his definition of immunity to error through wh-misidentification in terms of undermining defeaters is incorrect, and I argue that, on the correct definition, immunity to error through wh-misidentification is a status that bodily first-person judgments can have. The other attempt to undermine the Parity Thesis, discussed in section (§4), is offered by Coliva, and also Wright. They both focus on the notion of immunity to error through false-belief-or-false-background-presupposition misidentification. I argue that their definition picks out an epistemic status that no judgments have, so it definitely isn’t an epistemic status that some mental first-person judgments, but no bodily first-person judgments, have. Section (§5) concludes.
1. Immunity to error through false belief misidentification
To say that a judgment is immune to error through misidentification is to say that it is not possible for it to be in error as a result of a misidentification. The natural way of specifying a notion of IEM is therefore to say what the corresponding notion of misidentification is. In this, it is natural to be guided, at least provisionally, by the ordinary use of the term ‘identify’. That word has two main uses. On one of them, to identify an object is, roughly, to distinguish it from its surroundings – to single it out. For example, I might identify the tiger in the grass when my attention seizes on a striking orangey form moving amid the surrounding green. On a different use, to identify an object is, roughly, to see that it is identical to something else. For example, I might identity the tiger I have just spotted in the grass as Tony Tiger, a tiger I know from way back.
It is the second use of the term ‘identity’ that is most relevant for the first kind of IEM I want to introduce. Plausibly, an identification, in the second sense, is some kind of attitude with a content involving identity. A bit more tendentiously, but still plausibly, an identification is a belief with a content involving an identity. Suppose I judge ‘Tony Tiger is swishing his tail’. I may not have performed a conscious chain of reasoning involving the premises ‘That tiger is swishing his tail’ and ‘That tiger is Tony Tiger’. Recognition of Tony Tiger may have been instantaneous, so that I came by the suite of beliefs ‘That tiger is swishing his tail’, ‘That tiger is Tony Tiger’ and ‘Tony Tiger is swishing his tail’ in one fell swoop. Still, the first two beliefs are the basis for my third belief, the belief that Tony Tiger is swishing his tail. The first two beliefs causally sustain that third belief. And they are what I would appeal to if asked to give my reasons for coming to that belief.
One notion of IEM is the following:
A judgment ‘a is F’ is IEM, relative to basis b, if and only if b does not include an identity belief. (Immunity to error through false-belief misidentification)Footnote 5
An important thing to note for what follows is that, on this notion of IEM, a belief’s not being based on any other beliefs – i.e. being non-inferential – guarantees that it will be IEM. This is so, even if the belief in question is vulnerable to quasi-cases.
2. Quasi-cases
The most famous kind of quasi-case are cases of quasi-memory.
A subject quasi-remembers an event e iff e took place, the subject has an apparent memory of e, and this apparent memory causally derives from someone else’s past perceptions and experiences.
For example, imagine that two hemispheres of someone’s brain are separated and each is put in a new body. This produces two new people each of whom is psychologically continuous with the original person but neither of whom is identical with her. Because each of the new people has inherited half the original person’s brain they quasi-remember many events in the original person’s life. Say the original person was once on a ship. As a result, both of the new people may each now have a quasi-memory as of having been on a ship.
Quasi-memory seems to be a species of a broader genus. For example, we can say that someone quasi-proprioceives the instantiation of a bodily property (e.g. having crossed knees) iff that bodily property actually is instantiated, the subject has apparent proprioceptive awareness as of their instantiating that property, and this apparent awareness derives from the state of somebody else’s body. I might have apparent proprioceptive awareness as of my knees being crossed because I am appropriately hooked up to somebody else’s body, and that person’s knees are crossed. Or again, we can imagine a case of quasi-vision in which it visually appears to me that I am standing in front of a tree because I am hooked up to, and receiving visual information, through the eyes of somebody else who really is standing in front of a tree.
A ‘quasi-case’, we can say, is a case in which all these three elements figure. A subject has an impression that a property is instantiated. It seems to the subject as though they are the one who instantiates the property. In fact, it is because someone else instantiates the property that the subject has the impression they have.
It is very plausible that every bodily first-person judgment will be vulnerable to quasi-cases. The contentious question is whether quasi-cases involve misidentification, and whether a judgment’s being vulnerable to quasi-case therefore makes it vulnerable to error through misidentification. It cannot mean that, if our notion of a misidentification is just that of a false identity belief. ‘My knees are crossed’, based on proprioception, is vulnerable to quasi-proprioception. But it may still be based directly on experience and, in that case, it won’t be based on an identity belief. So the opponent of the Parity Thesis needs a different notion of IEM.
3. Immunity to error through wh-misidentification
There is a notion of misidentification that is different from the notion of a false identity belief. I will use Pryor’s term for this different kind of misidentification: wh-misidentification. The phenomenon of wh-misidentification can be explained in the abstract, and examples of it can be given, without mentioning quasi-cases. The question then arises of whether quasi-cases involve error through wh-misidentification. Pryor argues that the answer to this question is ‘Yes’.Footnote 6 I first rebut his argument, which depends on defining immunity to error through wh-misidentification in terms of undermining defeaters. I then give an argument that the answer is ‘No’. So, the conclusion of the discussion of Pryor will be that immunity to error through wh–misidentification is indeed a different, and more demanding, notion from immunity to error through false belief misidentification. But it is still a notion on which the Parity Thesis holds.
To get a grip on what wh-misidentification is notice that if a judgment of the form ‘a is F’ is in error it seems a reasonable and potentially instructive question to ask whether the error is due to the singular part of the judgment, or to the predicative part of the judgment. A symptom of the error being due to the predicative part of the judgment would be if there was another belief about the same object that the subject retains knowledge of. E.g., if I judge ‘That vase is red’, because I see a white vase bathed in red light, I may nevertheless retain knowledge of ‘That vase appears red’. A symptom of the error being due to the singular part of the judgment would be if there was another belief about the same property that the subject retains knowledge of. E.g. if I judge ‘Tony Tiger is swishing his tail’, because I see Tommy Tiger swishing his tail and I take Tommy for Tony, I may retain knowledge of ‘Something is swishing its tail’.
A perfectly good notion of ‘misidentification’ is just the notion of an error that in some way involves the singular component of the judgment. One species of this genus is where the subject falsely identifies one thing as a different thing, e.g. by forming an identity belief. But, it turns out, there is a different species of the same genus – wh-misidentification.
We encounter wh-misidentification in the following example of Pryor’s (Reference Pryor1999). In The Little Girl Case a subject is looking at a little girl who is standing 10 feet in front of them. The subject’s visual experience justifies them in judging ‘There is a little girl standing about 10 feet in front of me’. So the subject makes that judgment on a non-inferential basis. So, in particular, it is not based on an identity belief.
Suppose though that the subject is in a hall of mirrors, that there is a full-length mirror two feet in front of the subject, and that this mirror has interfered with the subject’s capacity to accurately locate the little girl. The little girl is really some distance behind the subject. The subject may still know, on the basis of vision, ‘There is a little girl somewhere or other in my vicinity’, which is the existential generalization of ‘There is a little girl standing about 10 feet in front of me’. We might describe the situation as follows: the subject has correctly detected a feature in their environment – the presence of a little girl – but they have gone wrong in identifying – or singling out – the location that instantiates that feature. But, to bring out the difference with the other species of misidentification, that isn’t because they have singled out a different location, and formed a belief that it is the location of the feature.Footnote 7
On Pryor’s way of thinking about memory, memory judgments are similar to the judgment ‘There is a little girl standing about 10 feet in front of me’ in The Little Girl Case. If I take at face value a memory impression as of having been on a ship that is in fact a quasi-memory, then I will have succeeded in detecting a feature of the world’s history – that someone was once on a ship – but I will just have gone wrong in identifying myself as the person who has that feature. So, memory judgments will be vulnerable to error through wh-misidentification.
To work out whether this is true we need to motivate a definition of immunity to error through wh-misidentification, and then apply the definition to memory. I first argue that Pryor’s definition, which is intended to yield the verdict that memory judgments are vulnerable to wh-misidentification, and which brings in the notion of an undermining defeater, is unacceptable. Then I motivate my own definition of immunity to error through wh-misidentification, which gives the opposite verdict. Footnote 8
(i) Pryor’s attempt to define immunity to error through wh-misidentification using undermining defeaters
Pryor’s first attempt at a definition says that that a judgment will be immune to error through wh-misidentification, relative to certain grounds, if and only if:
‘…there is no [proper] “part” of your justification for believing that a is F which could offer you knowledge that something is F, while leaving it an open question for you whether a is F.’ (Reference Pryor1999, 283)
Pryor is – admirably – not content to leave the notion of ‘part’ of your justification as an undefined primitive. He connects the notion with the possibility of one’s justification, or grounds, being defeated. The crudest articulation of the connection would be to suggest that a part of your grounds for p offers you knowledge that q if and only if your grounds for p could be defeated in such a way that they still offer you knowledge that q. But, as Pryor himself points out, this theory is too crude. For any p and q, and for any grounds, you can imagine God (or some other known-to-be-reliable giver of testimony) coming along and saying ‘p is false, but q isn’t’. Once you have the expanded set of grounds that include God’s testimony, you will be offered knowledge of q. This does nothing to show that part of your original grounds offer you knowledge of q.
The moral Pryor takes from this is that we need to invoke a distinction between two kinds of defeater. An undercutting defeater for p, Pryor says, undermines one’s justification for p without adding to one’s case for not-p. For example, if it seems to me that the wall is blue and you tell me that my eyes aren’t working well today, that is an undercutting defeater. An additive defeater for p, Pryor says, undermines my justification for p and adds to my case for not-p. If you tell me straight out that the wall is not blue, that is an additive defeater.Footnote 9 Pryor’s suggestion is this:
‘What we should say is this: “part” of your grounds justifies you in believing that p iff it is possible to take away some of those grounds, by undercutting them, in such a way that what is left will (still) justify you in believing that p.’ (Reference Pryor1999, 284)
Feeding Pryor’s understanding of ‘part of your grounds’ into his first attempt at a definition, we get Pryor’s finished definition.
A judgment ‘a is F’, believed on grounds g, is vulnerable to error through wh misidentification if and only if:
‘It is possible for g to be defeated by undercutting evidence in such a way that:
(i) the combination of g and that evidence no longer justifies you in believing that a is F, but
(ii) the combination of g and that evidence could, by itself, offer you knowledge that something is F. (Reference Pryor1999, 284)’Footnote 10
Obviously, a judgment is immune to error through wh-misidentification if and only if it is not vulnerable to wh-misidentification.
(ii) The argument against Pryor’s definition
Suppose ‘a is F’ is a judgment whose IEM status, relative to grounds g, is being adjudicated. Pryor’s definition tells us that we should ask whether there is a defeater of an appropriate kind – the undercutting kind – that can be added to g so that the subject is no longer justified in judging ‘a is F’ but is offered knowledge of ‘Something is F’. How exactly should we understand the restriction to undercutting defeaters? Is the restriction merely that the defeater not add to the subject’s case for ‘a is not F’ (this is all that Pryor’s own formulations, interpreted literally, require)? Or is it, more demandingly, that the defeater not add to the subject’s case for any other proposition, including, for example, ‘Something is F’?
Suppose that we interpret the restriction in the less demanding way. Then there is a quick argument that Pryor’s definition implies that any judgment of the form ‘a is F’ will be vulnerable to error through wh-misidentification. The relevant defeat scenario isn’t one in which God comes along and says to me: ‘Well, actually, no – a isn’t F. But I’ll tell you this much. Something is F’. Rather, God says: ‘Your judgment that a is F isn’t based on its being the case that a is F in the right way for that judgment to be knowledge. But I’ll tell you this much. Something is F’. In this kind of case, God’s testimony meets Pryor’s restriction, interpreted in the less demanding way. It does not add to my case that a is not F. The fact that my judgment isn’t based on its being the case that a is F in the right way for that judgment to be knowledge is entirely compatible with its being the case that a is F. But I do end up in a position in which I am offered knowledge of ‘Something is F’.
Suppose instead that we interpret the restriction in the more demanding way. Doing so will make it harder to craft a defeat scenario for a memory judgment that meets the restriction to undercutting defeaters. Consider as a first candidate a scenario in which a subject is told straight out that their memory impression as of having been on a ship is a quasi-memory. That is – once we unpack the notion of a quasi-memory – they are told that their memory impression (i) is not a memory of their own past but (ii) does causally derive from somebody else’s having been on a ship. This testimony entails, and so clearly adds to the subject’s case for, ‘Someone was on a ship’. So it is not an undercutting defeater, on the more demanding interpretation.
However, the defeat scenario Pryor focuses on is more subtle. It involves a victim of quasi-memory who is told:
‘… that some of his memories are quasi-memories, and that none of his memories as of being F [e.g. as of being on a ship – author] derive from derive from actual events in his own life. We leave it open whether his memories as of being F are among the quasi-memories of someone else’s past. Call this information D*. D* undercuts the justification that S’s quasi-memories give him for believing that he was F.’ (Reference Pryor1999, 295)
It’s a delicate matter whether D* really is undercutting in the more demanding sense. One might well insist that D* does add to the subject’s case for ‘Someone was on a ship’. The obvious difference from the scenario in which the subject is actually told outright that their memory is a quasi-memory is that D* (even if its accuracy is assumed) doesn’t provide a conclusive basis for ‘Someone was on a ship’. But it surely adds something to the subject’s case for ‘Someone was on a ship’, which makes it an additive defeater.
It’s also a delicate question whether the subject’s memory impression, supplemented with D*, offers them knowledge of ‘Someone was on a ship’. One might argue that, since the testimony in D* (even if accurate) is not a conclusive basis for ‘Someone was on a ship’, the subject does not end up with grounds that offer them knowledge of ‘Someone was on a ship’. Rather, they ought to regard the matter as left somewhat open.
These two worries should make us doubt whether Pryor has managed to describe a defeat scenario for a memory impression as of being on a ship that does all of following. First, it is undercutting (in the more demanding sense). Second, it undermines the subject’s justification for ‘I was on a ship’. Third, it leaves the subject with grounds that offer them knowledge of ‘Someone was on a ship’. But I do not want to insist on either worry.
What I would insist on rather is a conditional. If the scenario Pryor describes shows that memory judgments are vulnerable to error through wh-misidentification, then there will be analogous scenarios that reveal the same about any fallible judgment. So, granted that we don’t want the notion of IEM to collapse into that of infallibility, we must reject Pryor’s definition.
For any fallible judgment ‘a if F’, relative to any grounds g, we can distinguish two kinds of case in which the judgment is in error. One is a case in which something other than a really is F and that other thing’s being F is implicated in one’s judging ‘a is F’ (this is analogous to quasi-memory).Footnote 11 The other is a case in which there isn’t anything else whose being F is implicated in one’s judging ‘a is F’ (this is analogous to more ordinary kinds of memory failure). We can always imagine someone with some justification for ‘a is F’ being told that their judgment is not based on its being the case that a is F. It has gone wrong in either of the two distinguished ways, but we’re not saying which. The subject ends up with at least some justification for ‘Something is F’ but his justification for ‘a is F’ is completely undermined. If, in general, the possibility of this kind of defeat scenario is enough to show that ‘a is F’ is vulnerable to error through wh-misidentification, then it will follow that every fallible judgment is vulnerable to error through wh-misidentification. If it isn’t, then the kind of defeat scenario Pryor describes will not show that memory judgments are vulnerable to error through wh-misidentification.
There are quite a few delicate issues involved in working out which judgments satisfy Pryor’s finished definition of immunity to error through wh-misidentification. There is the initial exegetical issue about the more and less demanding interpretation of ‘undercutting’. Once that has been resolved (on the more demanding interpretation, most plausibly), there is a substantive question about which defeat scenarios count as undercutting, and another substantive question about which defeat scenarios leave the subject with grounds that offer them knowledge of the relevant existentially general proposition. So I have do not have settled view on what the extension of Pryor’s definition is. But I have argued that the only way its extension is going to turn out to exclude memory judgments is if it excludes every fallible judgment.
(iii) A definition in terms of independent knowledge
Rebutting Pryor’s argument that bodily first-person judgments are vulnerable to error through wh-misidentification isn’t the same as establishing that they are immune to it. To establish that they are we need to argue for, and then apply, an alternative definition.
The definition I propose is as follows:
A judgment of the form ‘a is F’ is, relative to certain grounds g, immune to error through wh-misidentification if and only if g justifies ‘a is F’ without offering independent knowledge of ‘Something is F’. (Immunity to error through wh-misidentification).
Pryor’s first attempt at a definition invoked the idea of one’s grounds dividing into different parts and some part of one’s grounds supporting ‘Something is F’ without supporting ‘a is F’. We have seen that this ‘part’-based approach goes astray when the notion of part is cashed out in terms of that of an undercutting defeater. I think we should focus on what seems to the root idea behind the part-based approach: the idea of independence. If some part of my grounds supports one proposition without supporting a different, logically stronger, proposition that can only be because my grounds give some support to the weaker proposition that is independent of the support they give to the stronger proposition.
Invoking independence will only help if we have a firm basis for judgments of independence. I propose the following test:
Grounds g offer knowledge of q that is independent of the justification they provide for p if and only if one can know that q on the basis of g, even if p is false (Independence Test).
What makes this test appropriate? A certain fact about knowledge: the fact that it is not possible to gain knowledge by inferring from false premises.Footnote 12 Suppose that ‘a is F’ is false but my grounds for ‘a is F’, g, nevertheless offer me knowledge of ‘Something is F’? Then it must be that g offer me knowledge of ‘Something is F’ that is independent of the justification they provide for ‘a is F’. If g supported ‘Something is F’ only by supporting ‘a is F’, from which ‘Something is F’ can be inferred, the fact the ‘a is F’ is false would deprive me of knowledge of ‘Something is F’.Footnote 13
The definition I have argued for returns the intuitively correct verdict in The Little Girl Case. The subject’s ‘There is a little girl 10 feet in front of me’ judgment is false. But the subject can still know ‘There is a little girl standing somewhere or other in my vicinity’. It follows that the subject’s grounds must offer independent knowledge of ‘There is a little girl standing somewhere or other in my vicinity’. So, on my definition, the judgment will count as vulnerable to error through wh-misidentification, as it intuitively ought to.
The verdict my definition returns about the ‘I was on a ship’ judgment, based on a memory impression as of having been on a ship, is that it will be immune to error through wh-misidentification. For the memory impression does not offer knowledge of ‘Someone was on a ship’ that is independent of the justification it provides for ‘I was on a ship’. If it did offer such independent knowledge, then the subject could know ‘Someone was on a ship’, just on the basis of the memory impression, even if ‘I was on a ship’ were false. But, as Evans persuasively argues, that isn’t the case.
Evans first points out that a victim of quasi-memory is unlikely to spontaneously come up with the hypothesis that they are a victim of quasi-memory. He then adds that, even if they did, their hypothesis would not be one that their memory impression itself justifies. So, the subject might judge ‘Someone was on a ship’, just on the unjustified suspicion that their memory impression is a quasi-memory. But in that case their judgment would merely be a (successful) ‘shot in the dark’ – true but not knowledge.
Evans’s assessment here has enough intuitive plausibility that Pryor concedes it.Footnote 14, Footnote 15 Having conceded it, Pryor is forced to argue for a definition of immunity to error through wh-misidentification on which the crucial issue is what knowledge the subject is offered when their memory impression is appropriately supplemented (by undermining defeaters). The core of my argument that memory judgments are immune to wh-misidentification has therefore consisted in an argument that the correct definition of immunity to wh-misidentification does not mention defeaters. Once the alternative, defeater-free definition is accepted, it is a fairly immediate consequence that the memory judgment ‘I was on a ship’ – despite being vulnerable to quasi-memory – will be immune to error through wh-misidentification. Similarly, other non-inferential first-person bodily judgments – such as those based on based on proprioception or perception – turn out to be immune to error through wh-misidentification, despite being vulnerable to quasi-cases.
I suggested at the beginning of this paper that it was the conjunct of the Parity Thesis that says that bodily first-person judgments can be IEM that is most controversial. I took for granted the other conjunct that says there are some mental first-person judgments – in particular, those based on introspection – that are IEM.
Now that we have in place a positive definition of immunity to error through wh-misidentification, it is worth testing it against the minority strand in the literature on IEM that claims that even introspective judgments fail to be IEM. Campbell (Reference Campbell1999b) initiated this strand, with his appeal to the phenomenon of schizophrenic patients who introspect their own thoughts, but who think those thoughts are someone else’s – so called cases of thought insertion. A typical example is one in which a subject introspects the thought ‘I am thinking p’ but is willing only to endorse ‘Someone is thinking p’ or ‘That other person is thinking p’. This kind of example fails to supports the verdict that introspective judgments are vulnerable to error through wh-misidentification, on the definition I have argued for. On that definition, what is required would be a case in which the subject is offered knowledge of ‘Someone is thinking p’ even though ‘I am thinking p’ is false. But the thought-insertion cases are cases in which ‘I am thinking p’ is true, not false. Footnote 16
In summary, I’ve argued that immunity to error through wh-misidentification really is different from the notion of immunity to error through false belief misidentification. The former is the more demanding notion so, as between the two, we should prefer to read the Parity Thesis as concerning the former. However, even on that reading, the Parity Thesis is true.
4. Immunity to error through false belief-or-background-presupposition misidentification
My notion of immunity to error through false belief misidentification mentioned the basis for one’s belief. My notion of immunity to error through wh-misidentification, as well as Pryor’s definition of immunity to error through wh-misidentification, mentioned the grounds for one’s belief. I didn’t highlight this distinction, since it was possible to give my argument against Pryor’s definition, and for my own, while ignoring it. In the definition of IEM under discussion in this section, however, the distinction will take center-stage.
Earlier, I provided a gloss on the notion of a ‘basis’ for belief. To bring out the difference with the notion of ‘grounds’ for belief, consider the following case, discussed by Pryor at (Reference Pryor1999, 276). I have justification for ‘The gas meter reads “E”’ and, because of that, I also have justification for ‘I am out of gas’. Because of fears about an evil demon, I form neither belief. So, it is not the case that I have an ‘I am out of gas’ belief that is based on a ‘The gas meter reads “E”’ belief. Even so, there is a kind of epistemic dependence here. The justification I have for ‘I am out of gas’ depends on the justification I have for ‘The gas meter reads “E”’. One can use ‘grounds’ as a label for this relation of epistemic dependence. In this case, we can say that ‘The gas meter reads “E”’ is included in my grounds for ‘I am out of gas’.
One might expect that in cases in which a subject actually does form a belief, the two notions – bases and grounds – will march fairly closely in step. One might think that, if a subject is justified in believing p, and the grounds for p include q, then their belief in p will have to be based on belief in q. One might also think the converse of this: if a subject is justified in believing p, and their belief in p is not based on belief in q, then it cannot be that their grounds for p include q. Applying this to the case of memory, one might argue that since a subject may be justified in judging ‘I was on a ship’ just on the basis of a memory impression, their grounds for that judgment cannot include the identity ‘I am the person from whose past my memory impression derives’.
The advocate of the notion of IEM that I discuss in this section says that this is too hasty. They think that ‘I was on a ship’, even if just based on a memory impression, will include that identity in its grounds. A subject can form a judgment that is justified, where the grounds on which the justification depends include a claim that is not part of the basis on which the subject relied. They can if that claim is a background presupposition.
The definition of IEM that this leads them to canvass is as follows:
A judgment ‘a is F’ is, relative to certain grounds g, IEM if and only if g (i) do not include an identity-belief and (ii) do not include a background presupposition of identity. (Immunity to error through false-belief-or-background-presupposition misidentification)
This kind of definition was first set out in Coliva (Reference Coliva2006) and followed up in Coliva (Reference Coliva, Evans, Prosser and Recanati2012). It has also been explored in Wright (Reference Wright, Prosser and Recanati2012). There are some differences in formulation between Coliva and Wright, but none that make a difference to my argument. I shall target the proposal generically, while also noting a few small differences in the two presentations.
How are we to understand the notion of a background presupposition? Background presuppositions are supposed to be a kind of epistemic ground, so they must be epistemically relevant to the thing being judged. The relevance, however, isn’t foregrounded in the subject’s own psychology: it is in the background. I’ll explain these two elements in reverse order.
There are in fact two different ways that a ground’s relevance could be foregrounded in the subject’s own psychology and that background presuppositions therefore contrast with. I’ll illustrate them using the case discussed at Coliva (Reference Coliva, Evans, Prosser and Recanati2012, 417). Suppose I see a woman, my aunt Miriam, and form two separate judgments, ‘That woman is wearing a red hat’ and ‘That woman is Aunt Miriam’, from which I consciously infer ‘Aunt Miriam is wearing a red hat’. There is dependence on an identity here, of an extremely foregrounded variety.
Here is a slightly different possibility. I see Aunt Miriam. I don’t form two separate judgments, one of which is an identity, and consciously infer ‘Aunt Miriam is wearing a red hat’. I come by my ‘That woman is wearing a hat’ and ‘Aunt Miriam is wearing a hat’ beliefs in one fell swoop. Still, if I were asked to justify my ‘Aunt Miriam is wearing a red hat’, I would appeal to the identity. I might say, for example, ‘It’s perfectly clear that the woman I’m looking at is wearing a red hat. And I can see that that woman is my aunt Miriam’. There is dependence on an identity here, of an at least somewhat foregrounded variety.
If either of these kinds of dependence is present, then ‘That woman is aunt Miriam’ is part of the basis on which ‘Aunt Miriam is wearing a hat’ is judged and is not a mere background presupposition. However, notice that in the case of the memory judgment ‘I was on a ship’ neither kind of dependence on an identity seems present. I definitely don’t consciously infer ‘I was on a ship’ from ‘I am having a memory impression as of being on a ship’ and ‘I am the one from whose past this memory impression derives’. In addition, it’s very unlikely that that if someone challenged me to defend my ‘I was on a ship’ judgment, I would offer those premises. So, ‘I am the one from whose past this memory impression derives’ is not part of the basis I rely on.
That is why it is important for Coliva and Wright that there is a third kind of ground – background presuppositions. The test for a content’s being a background presupposition is, first, that a belief in it isn’t part of the basis the subject relies on (if it were, the dependence would be foregrounded in the subject’s own psychology) and second, that a positive condition is met, whose obtaining captures why the content is nevertheless relevant. Coliva and Wright state the positive condition slightly differently.
For Coliva (see Reference Coliva, Evans, Prosser and Recanati2012, 417):
p is a background presupposition of q only if: were p to fall into question, a (rational and fully conceptually equipped subject) would be prepared to withdraw from q.
For Wright (see Reference Wright, Prosser and Recanati2012, 269) (at least for judgments based directly on experience, which are the kind of judgment we are interested in):
p is a background presupposition of q only if: if the subject doubted p, they would doubt that their experience justified q.
One can have more that one set of grounds for a single judgment. So I take it that Coliva’s ‘withdraw from q’ is implicitly relativized to a set of grounds. E.g. if I thought that my memory was utterly unreliable, I might withdraw from judging ‘I was on a ship’ on the grounds of memory but still be reluctant to withdraw from ‘I was on a ship’ tout court, because I think that most people have been on ships at one point or another, and don’t see why I should be any exception. Wright achieves the same kind of relativization by talking not about withdrawing from q but about doubting that one’s experience justifies q. I will take the relativization as read.
Coliva specifies that our subject is rational and fully conceptual equipped, which Wright omits to do. I take it though that Wright should be read as assuming that his subject satisfies something like Coliva’s restriction. It’s utterly unpredictable what a sufficiently irrational subject would do, if they doubted p.
Suppose I judge, on the basis of a memory impression, ‘I was on a ship’. Someone comes along and queries my ‘I was on a ship’ judgment. I very likely won’t immediately start thinking about quasi-memory involving possibilities. The person who is putting me under pressure didn’t specifically mention that kind of possibility, and it is far from the first thing anyone would think of unprompted.
Something different could happen though. Someone could come along and zero in on that possibility themselves. Perhaps they don’t even mention my ‘I was on a ship’ judgment. They simply query whether, as they would put it, ‘You are the person from whose past your memory impression derives’. I certainly wasn’t thinking about that possibility before. But I can hardly claim that the question is irrelevant to my ‘I was on a ship’ judgment, which is based on memory. So if their query leads the identity to seriously fall into question for me in the sense that I actually start to doubt it, I surely had better retreat from ‘I was on a ship’. So, the identity does meet Coliva and Wright’s test for being a background presupposition of ‘I was on a ship’. If having an identity as a background presupposition makes for vulnerability to misidentification, as it does on the definition of IEM they propose, then ‘I was on a ship’, based on memory, is not IEM after all.
Coliva draws this conclusion in relation to her example of a memory judgment:
So the presence of identification components featuring merely as background presuppositions allows for the possibility of judgments – like “I was in that remote part of Scotland five years ago” when based on one’s memories – which, despite being grounded in irreducibly first personal contents, and having no identification component among their grounds at all from the subject’s point of view, may in principle presuppose a false identification (Reference Coliva2006, 419)Footnote 17
Wright draws a similar conclusion in relation to proprioceptive judgments:
‘Consider the proprioceptively based judgment that my legs are crossed. …this judgment is grounded in a pure experience … But however that may be, the judgment does nevertheless rest on an identification, namely that it is my body – my legs – that are the source of the propriocetive sensations that I am having – or perhaps better: the person whose arrangement of limbs is the causal source of my current proprioceptive experience is myself ….One consequence is that such judgments are not IEM.’ (Reference Wright, Prosser and Recanati2012, 271)
Coliva and Wright’s focus on background presuppositions might initially appear like an ingenious way of capturing a dependence of a particularly subtle kind on an identity. In fact though, the strategy overshoots. The attempt to craft a definition of IEM that rules out memory or proprioceptive judgments as IEM results in a notion of IEM that captures an uninteresting epistemic status, one that no judgments have. The root problem is that background presuppositions are so promiscuous.
Take an arbitrary judgment of the form ‘I am F’.Footnote 18 And now consider a scenario in which I come to doubt whether I really am the person whose being F is responsible for my judging that I am F. In that scenario, I would of course come to doubt whether I am justified in judging ‘I am F’. It follows that ‘I really am the person whose being F is responsible for my judging that I am F’, an identity, is a background presupposition of my judgment. Since ‘I am F’ was arbitrarily chosen it follows that, on this definition, no judgments are IEM.
That argument seems very quick. It might be objected that, for some values of F, there will be no possible scenario in which I could be made to doubt whether I really am the person whose being F is responsible for my judging that I am F.
This seems wrong. It is certainly wrong for fallible judgments and it also seems wrong for infallible judgments. I will take the two cases separately.
Suppose ‘I am F’ is fallible. Then there will be possible cases in which I judge ‘I am F’ even though I am not F. If there are such cases, then a subset of the cases will be ones in which there is someone else who really is F, and where that person’s being F contributes to bringing about whatever combination of circumstances lead me to judge ‘I am F’. If such cases are possible, then there is no reason why I couldn’t be given reason to suspect that they obtain. So, no fallible judgment will count as IEM.
Perhaps surprisingly, the same is true for infallible judgments, if there are any of those. The key point is that even if a judgment is infallible, I may be persuaded that it is fallible. E.g. even if ‘I am in pain’, based on introspection, is infallible, I may be convinced by clever but ultimately unsound anti-luminosity arguments that it is fallible. I may further be convinced that, in a particular case, my inclination to judge ‘I am in pain’ is in error and moreover there is someone else whose being in pain has contributed to bringing about my inclination to judge that I am in pain. E.g. suppose I am convinced that there is someone else whose being in pain regularly causes me to have an itch that is nearly but not quite intense enough to count as a pain, and that I am apt to mistake for a pain. I am further convinced, on a given occasion, that this has actually happened. In that case, I ought to withdraw from ‘I am in pain’ because of my doubt about the identity ‘I am the person whose being in pain is responsible for causing me to be inclined to judge “I am in pain”’. This is despite the fact that, we are supposing, ‘I am in pain’ is in fact infallible. So, even infallible judgments will count as vulnerable to error through misidentification, on the definition being considered. That definition is unacceptable as a definition of IEM, or as a way of fleshing out the notion of thought about oneself as subject, because it captures an epistemic status that none of our judgments have.
One might offer a test for being a background presupposition stronger than the conditional one that I, following Coliva and Wright, have applied. But, first, if advocates of the definition had an appropriate stronger test up their sleeve, they presumably would have offered it. Second, and more importantly, Coliva and Wright’s arguments that memory and proprioceptive judgments have identities as background presuppositions did rely on the conditional test for being a background presupposition being the right test. We can strengthen the notion of background presupposition to make it less promiscuous if we like. But we have no reason to expect that, on the new notion this yields, first-person bodily judgments will still count as always involving background presuppositions of identity.
5. Conclusion
I’ve discussed three different epistemic statuses – immunity to error through false belief misidentification, immunity to error through wh-misidentification, and immunity to error through false belief-or-background-presupposition misidentification.
In relation to the first status, immunity to error through false belief misidentification, I’ve argued, in agreement with everyone else in the literature, that bodily first-person judgments can have it.
In relation to the second status, immunity to error through wh-misidentification, I’ve argued that it really is a different, and more demanding, status than immunity to error through false belief misidentification (agreeing with Pryor). But I’ve argued that it is a status that bodily first-person judgments can have (disagreeing with Pryor). They won’t have it, if we understand immunity to error through wh-misidentification in terms of undermining defeaters, as Pryor proposes we do. But we shouldn’t do that, since doing that causes the notion of immunity to error through wh-misidentification to collapse into that of infallibility.
In relation to the third status, immunity to error through false belief-or-background-presupposition misidentification, I’ve argued that bodily first-person judgments cannot have it (agreeing with Coliva and Wright). But that is only because the third status turns out to be one that no judgments have. We might avoid that undesirable result by modifying the definition of what it is for something to be a background presupposition. But we have no reason to expect that, on the new definition, first-person bodily judgments will still count as always involving background presuppositions of identity.
My conclusion is that we have no evidence against the Parity Thesis. On each of the two notions of IEM that succeeds in capturing an interesting epistemic status, immunity to error through false belief misidentification and immunity to error through wh–misidentification, the Parity Thesis is true. Of course, it doesn’t follow that we should endorse the Parity Thesis. There might turn out to be some interesting epistemic status that some mental first-person judgments have, that no bodily first-person judgments have, and that relates sufficiently closely to the kind of examples used to introduce the idea of immunity to error through misidentification that that status would count as a new kind of immunity to error through misidentification. If one thinks that there must be such a status, then the moral of this paper is just that we need to keep on looking for it. But the failure to find it, after considerable effort, naturally makes pressing the question of what reason there is to think that there must be such a status. The alternative possibility is that when I think about, e.g., the orientation of my limbs, or my location relative to the things in my environment, or my skin being covered by small beads of sweat, I think about myself as a bodily thing, and simultaneously as a subject – so I think about myself as a bodily subject. I’ve argued that the discussion of immunity to error through misidentification, and in particular of quasi-cases, does nothing to undermine this possibility. Is there any independent reason to assign it a low probability?Footnote 19
Disclosure statement
No potential conflict of interest was reported by the author.
